JP2580566B2 - Speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Speech data including a character string indicating a phrase to be synthesized and pronunciation information for specifying what kind of speech is to be synthesized with the character string is stored in a key corresponding to each audio data. A voice information dictionary stored with the voice information dictionary is provided, a desired voice data is read out by designating a key from the voice information dictionary, and the read voice data is supplied to voice synthesis means, thereby synthesizing a desired voice. .

[Industrial applications]

The present invention relates to a speech synthesizer that can obtain desired speech data from a speech dictionary created in advance by designating keys instead of individually specifying pronunciation information in speech synthesis.

[Prior Art] Recent advances in speech synthesis technology have made it possible to use speech synthesis means, such as a speech synthesis card, even in personal computers (hereinafter abbreviated as personal computers), etc., making it relatively easy to use speech. , And various systems using voice response have been developed accordingly.

In order to perform speech synthesis, it is necessary to provide speech data including a character string and pronunciation information to the speech synthesis means by some means or method.
The SIC language now supports TALK sentences.

An example in which audio data is specified in such a BASIC language will be described below.

As an example, in the case of uttering “Yamagata”, as the voice data to be specified, it is necessary to specify a character string indicating a phrase to be uttered, and subsequently, pronunciation information such as an accent position and a voice pitch. For example, to make the above “Yamagata” uttered with an accent on the second beat, it is described as TALK “Yamagata: A2”. 'Yamagata' in the above TALK sentence
Is a character string specifying the clause to be composed, followed by 'A
"2" is pronunciation information specifying that an accent is placed on the second beat, and ":" is a delimiter.

To realize a system using speech synthesis, a character string indicating the phrase to be uttered, the accent position, the intonation of the beginning or end of the word,
It is necessary to describe pronunciation information that specifies the pitch to be uttered, the pronunciation speed, the division between female voice and male voice, and the like. When the program is executed, these character strings and pronunciation information are transferred to a voice synthesis card, and the voice synthesis is performed. Done.

[Problems to be solved by the invention]

However, the description of the pronunciation information itself requires complicated work, and even in the same phrase, the pronunciation information is not constant depending on time and occasion, so it is necessary to specify desired pronunciation information each time. Not only does it require a great deal of man-hours, but there is also a problem that the specified result cannot be heard and confirmed on the spot. In order to hear the actual utterance, it is necessary to execute the TALK sentences included in the program one by one, which is a complicated operation. Therefore, it is not always easy to create a program that uses speech synthesis.

Therefore, an object of the present invention is to simplify designation of a desired phrase and desired pronunciation information.

[Means for solving the problem]

In order to achieve the above object, the present invention provides a voice information dictionary storing various voice data together with corresponding keys, searches the voice information dictionary by specifying the keys, reads desired voice data, and reads the voice data. It was given to the synthesis means.

[Action]

Since various voice data are stored in the voice information dictionary together with corresponding keys according to a predetermined description format, by searching for the voice information dictionary by specifying a key,
Desired audio data can be easily obtained. Therefore, in a program that uses voice synthesis, it is only necessary to provide the voice data obtained by designating the key as it is to the voice synthesis means, which makes it very easy to create a processing program.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram for explaining the configuration of a speech synthesis system according to the present invention, wherein 1 is an operating system, 2 is speech synthesis means, and 3, 3 'are speech information dictionaries provided in the present invention (hereinafter simply referred to as speech dictionaries). 4 is audio data, 5
Is a key, and 6 is a processing program. 7 is a console,
8 denotes a direct access volume (DASD), and 9 denotes a floppy disk drive (FPD). Note that, as indicated by reference numeral 3 'in the figure, a speech dictionary is usually present on DASD8 or FPD9. The voice dictionary 3 shows this in an enlarged manner.

As shown in FIG. 2, the voice dictionary 3 stores a character string 11 indicating a phrase to be voice-synthesized and voice data 4 including pronunciation information 12 together with a corresponding key 5.

The pronunciation information shown in the figure is an example in which various parameters are described by one letter each of English letters and numerals, and the letters E, A, B,
I, L, and R respectively describe the change in the tone end point, the accent position, the change in the ending intonation, the pitch to be uttered, and the pronunciation speed by the numbers that follow.

The character strings corresponding to the keys "A0001" and "A0002" in the same figure all indicate the same phrase "Konichiha", that is, "Today is today", and specify how to utter this with each pronunciation information. doing. In the example of the figure, only the accent position is different, and the key “A0001” is in the “d” of the third beat,
'A0002' specifies that an accent is to be placed on the first beat of the "ko".

As described above, the voice dictionary 3 is provided not only with the case where the phrases are different but also with the case where the phrase is the same but the pronunciation information is different, and the keys indicating the respective words are provided. It is stored in the dictionary 3 as a random access file.

When a system using voice synthesis using such a voice dictionary, for example, a business system using voice response, is constructed, it is necessary to be able to search the voice dictionary at high speed. If the dictionary search is slow, problems such as slow response by voice occur.

When high-speed retrieval is required, an index sequential file is generally configured. However, a personal computer usually does not provide a function for constructing an index sequential file, and can use only a sequential file and a random access file. The former has a simple structure, but has a slow search, and the latter has a property that a high-speed search is possible if the record number is known.

When specifying voice data in a voice dictionary, it is generally unknown which record the data is,
It is impossible to search by specifying the record number.
In addition, in order to facilitate designation of voice data, it is desirable to use a symbol or word having some relation with the voice data to be searched as a key.

Therefore, a voice dictionary used on a personal computer must be capable of specifying voice data with a key such as a related symbol or word, and must be configured to enable high-speed search using this key. .

The reason why the key 5 corresponding to the voice data 4 is provided in the voice dictionary 3 and the voice data 4 is stored together with the key 5 is as described above.

If the voice dictionary 3 is configured as described above, the key 5
, A high-speed search can be realized by various methods. For example,
An index file is created using the key 5, the record number for the specified key 5 is known from the index file, and the record of this number is stored in the audio dictionary 3.
There is a method of retrieving the voice data 4 and reading the voice data 4.

Next, an example using the voice dictionary 3 configured as described above will be described.

In a processing program 6 for a business or the like using a voice response, the voice data 4
Is read out and passed to the speech synthesis means 2.

Suppose now that the phrase "today" is to be uttered with an accent on the first beat, the key "A0002" corresponding to the second voice data in FIG. 2 is designated and the voice dictionary 3 is searched. Then, it is specified in the processing program 6 to read the corresponding audio data 4. This operation can be performed, for example, by using an index file as described above. As a result, a desired character string 11 and pronunciation information 12 are obtained.

Since the character string 11 and the pronunciation information 12 obtained in this way follow a predetermined description format, they are
It is given to the speech synthesizing means 2 according to a usual method such as substitution into a TALK sentence. By executing the processing program 6 created as described above, the phrase “Today is”
It can be uttered as desired.

As described above, in the present embodiment, the processing program 6
There is no need to describe the necessary character strings 11 and pronunciation information 12 one by one, and a search is made in the audio dictionary 3 by specifying the key 5 corresponding to the desired audio data 4. 12
Need only be given to the speech synthesis means 2. Therefore, it becomes easy to create the processing program 6 using the speech synthesis.

In the above description, an example is shown in which the key 5 is composed of one English character and four numbers, but the configuration of the key 5 is not limited to this, and has some relation with the audio data 4. You may comprise so that it may make it. For example, it is also possible to use a combination of kana characters and numbers, such as "con 01" and "con 02" so as to easily associate phrases. In order to make the search of the voice dictionary 3 easy, it is necessary to search the records in the voice dictionary 3 using the key 5 according to a certain rule as described above. Therefore, the key 5 only needs to be suitable for determining the logical order of the record.
Furthermore, it is sufficient that the audio data 4 is configured to be convenient when it is selected.

The pronunciation information 12 shown in FIG. 2 is one example, and the designation method and type are not limited to the example shown in FIG. For example, as the pronunciation information, in addition to the six pieces shown in the figure, there is a distinction between a male voice and a female voice, the amount of accent enhancement, and the like, and the voice dictionary 3 may include these designations.

This speech dictionary 3 is usually stored in DASD 8 or FPD 9, but in order to speed up the retrieval of speech data when using the speech dictionary, it is desirable to place it on the main storage device of the speech synthesis system. However, unless the voice dictionary 3 is small, the voice dictionary 3 on the DASD 8 or FPD 9 is normally used as it is. In particular, this method is used when the voice dictionary 3 is large and it is difficult to arrange it on the main storage device, or when the search speed is not a problem.

The voice dictionary 3 is placed on the DASD 8 or FPD 9, a search is created using the key 5, the search is placed on the main storage device, and voice data on the DASD 8 or FPD 9 is searched by referring to this index. It is also possible to use the method.

As described above, the location of the voice dictionary 3 and the search method can be variously selected.

In this embodiment, for convenience of explanation, the length of the character string 11 is 5
Although the example of the character is shown, the length of the character string 11 is not limited to this, and it can be arbitrarily selected, and the pronunciation information 12 can specify various types other than those shown. Need not be particularly described.

〔The invention's effect〕

As described above, according to the present invention, it is not necessary to specify pronunciation information that requires complicated operations, and it is possible to specify audio data by simply selecting a key,
It is easy to create a processing program that uses speech synthesis.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the configuration of the present invention, and FIG. 2 is a diagram illustrating the configuration of a voice information dictionary. In the figure, 2 is a voice synthesizing means, 3, 3 'are voice information dictionaries, 4 is voice data, 5 is a key, 6 is a processing program,
11 indicates a character string, and 12 indicates pronunciation information.

Claims (1)

(57) [Claims] 1. A speech synthesizer, wherein the speech synthesizer has a character string indicating a phrase to be speech-synthesized, an accent position with respect to the character string, intonation of the beginning or end of a word, a pitch to be uttered, a pronunciation speed, a female voice or a male voice. In a speech synthesis apparatus that performs speech synthesis by giving speech data composed of pronunciation information such as classification of a speech, for each of a plurality of speech data corresponding to each of character strings indicating a phrase to be speech-synthesized, A voice information dictionary is provided for adding a corresponding key and storing the voice data together with the key. At the time of voice synthesis, the voice information dictionary is searched using a key indicating desired voice data, and the corresponding voice data is searched for. A voice synthesizing device wherein the read voice data is supplied to the voice synthesizing means.